Managing drought-induced mortality in Pinus radiata plantations under climate change conditions: A local approach using digital camera data

Abstract

Pinus radiata D. Don is an internationally important plantation species. The predicted increases in the frequency, duration and/or severity of drought and heat stress associated with climate change could challenge the commercial viability of P. radiata plantations in several countries where existing plantations are grown in districts already vulnerable to periods of drought. We undertook an empirical approach to examine silvicultural and site factors associated with a significant drought-induced mortality event in southern New South Wales, Australia. The aim was to identify local, practical, risk prevention options for P. radiata plantations exposed to the climatic conditions predicted for this region. Our approach did not rely on ground-based assessment of plots but rather a total census of dead trees across two study areas totalling 10,000 ha of P. radiata, in compartments ranging in age from 0 to 35+ years. Dead tree density counts were derived both manually and automatically from high spatial resolution digital multispectral imagery acquired using a Leica ADS40 Airborne Digital Sensor. The results showed a strong correlation between dead tree densities obtained by the manual detection and the automated process ( R = 0.95, P < 0.0001). Two modelling approaches were applied: random forests (RF) and generalised additive models (GAM). For our study sites, both methodologies identified a similar set of parameters, with time since planting in unthinned stands being the most influential variable, and terrain variables playing a smaller role. Specifically, the models identified a threshold age at around 17–18 years for stands on good quality sites and before age 16 on poorer quality sites before the on-set of catastrophic mortality under severe drought conditions. Both modelling techniques also identified similar trends with respect to elevation and slope, and indirectly, site quality. These site attributes being likely to contribute to water availability. If stands in this area are to be planted at 1000 stems ha −1 and thinning schedules cannot be met, then we recommend avoiding sites having a mean elevation of below 600 m or to establish these sites with drought-tolerant genotypes. These recommendations cannot be extrapolated beyond the range of the data. However, the application of robust image classification techniques (e.g. automated tree counts) to high spatial resolution digital imagery, the increasing availability of high resolution climatic, terrain and edaphic GIS datasets, as well as readily available spatial modelling packages, helps off-set the limitations in transferability of our empirical approach.